Challenging the limits of logP prediction accuracy: Results of the SAMPL6 blind challenge.
パトコア
The SAMPL (Statistical Assessment of the Modeling of Proteins and Ligands) challenge aims to evaluate the accuracy of biomolecular and physical modeling for rational drug design. The recently announced SAMPL6 assessment focused on the prediction of the octanol-water partition coefficient (logP). In this blind challenge, 91 predictions were submitted from 17 research groups for 11 compounds, utilizing quantum mechanics, molecular mechanics, knowledge-based, empirical, and hybrid methods. Among these, highly accurate methods were identified, with RMSE remaining below 0.5 logP units for 10 different methods. Inspired by these results, we verified the accuracy of ChemAxon's logP prediction tool. As a result, only one case (SM11) out of 11 had an absolute error exceeding 0.5, which was found to be the largest average error among empirical methods. The calculations of ChemAxon logP demonstrated high accuracy in predictions, suggesting that this model can contribute to the optimization of new molecules or experimental conditions throughout drug discovery projects. For more details, please see the link.
Related Links